SolSpotPerpFundingCarryNeutralBinance8H

SOL Cross-Venue Funding Carry, Market-Neutral (Long SOLUSDT.BINANCE_SPOT + Short SOLUSDT.BINANCE Perp on Extreme Positive Funding, Collect Funding While Delta-Neutral, Basis-Stopped, Multi-Day Hold, 3-Parameter)

A MARKET-NEUTRAL, cross-venue funding-carry strategy that is the SOL sibling of the factory's single highest-Sharpe deployed survivor (the BTC long-spot/short-perp funding carry, Sharpe 8.6) — applied deliberately to a DIFFERENT, structurally higher-funding asset. It fills two of the thinnest portfolio dimensions at once: cross_venue (7.3% vs >=15% target) and market-neutral direction, while leaning into the ONE mechanism class the factory has repeatedly promoted (funding carry: the BTC cross-venue carry AND the 5-asset negative-funding basket both reached paper). It is NOT a funding-DIRECTION signal on majors (L13's exhausted class — this does not bet on price, it harvests a mechanical cash flow while delta-neutral) and NOT a naked basis trade (L1 — it is explicitly gated to only trade when the funding accrual exceeds the round-trip cost, with a stated median). The asset choice is the edge differentiator: SOL perpetual funding is structurally higher and spikes far more often and more extremely than BTC's (routine +0.05% to +0.10% per 8h in bull phases = ~55-110% annualized), so the same delta-neutral mechanism collects materially more carry per episode than the BTC version. Only 3 parameters (funding entry threshold, funding exit threshold, basis stop) to stay clear of the overfit graveyard.

Iteration 3 fixes the Layer-4 fee-viability/carry-accumulation defect with the smallest possible change: only numeric parameter defaults move; all signal/entry/exit/sizing STRUCTURE that already passed Layers 1-3 is untouched (no import, class, or logic changes, so earlier layers stay green). The QA root cause was the exit band [enter +0.5bp, exit <=0], which round-tripped the book at every funding zero-crossing (~4 periods held, ~0.2bp carry banked against a ~30bp round-trip; turnover 77, total_return -9.7%). The mechanism the hypothesis describes requires a multi-day hold that amortizes the four-fill round-trip over many 8h funding periods. The fix is deep hysteresis plus a minimum hold: (1) entry_funding 0.00005->0.0001 so we initiate only in a genuine carry regime rather than on near-zero noise; SOL's median 8h funding is ~1bp so a +1bp gate still fires frequently, preserving the Layer-3 trade generation that already passed at 0.5bp. (2) exit_funding 0.0->-0.0001 so a position HOLDS THROUGH zero-crossings and is flattened only when funding clearly inverts to <=-1bp; this directly removes the zero-crossing churn the QA flagged and lets one position ride the whole positive-funding regime, accumulating carry across many periods. (3) min_hold_bars 1->3 so the book never round-trips inside ~1 day. The resulting wide [+1bp, -1bp] band collapses turnover (cutting the fee/impact drag the QA cited) and stretches holds so accrued funding clears the round-trip, moving avg_trade_return_pct off the ~0.2bp floor. Parameter count stays at three core knobs (entry_funding, exit_funding, basis_stop_pct); exit_funding simply now takes a negative value, which is a value change, not a new parameter. The strategy remains the delta-neutral, positive-funding-only carry the hypothesis specifies, on the correct BINANCE_SPOT + BINANCE cross-venue book.

Negative-expectancy delta-neutral SOL spot-perp funding carry — not worth 2 hours of optimization. Funding IS credited by the engine (current ground truth), so this is a real economic result, not an uncreditable-carry artifact: avg_trade_return_pct is -0.211% (NEGATIVE), profit_factor 0.95, expectancy -$220/trade, Sharpe -5.19 (CI [-7.13, -3.01]), PSR 0.0. The SOL funding collected while delta-neutral does not cover the two-leg (4-fill) round-trip commissions plus spot-perp basis tracking error — the identical outcome to the BTC spot-perp funding-carry sibling abandoned earlier today. The sample is also tiny and unmeasurable: despite a 2020-2026 SOL span, only 332 bars were jointly processed and all 23 trades (12 entries) fall in 2020-09→2021-01 (data_days 111, annual_returns only 2020/2021), which cannot populate a 3-window walk-forward plus holdout. benchmark_meaningful is correctly false (delta-neutral); on absolute metrics the book loses in every vol regime. The 'SOL funding is structurally higher than BTC' differentiator did not produce a fee-clearing edge, and no tuning of entry_funding/exit_funding/basis_stop flips a carry that is negative net of costs. Failure pattern: fee_edge/no_edge delta-neutral funding carry (carry < round-trip + basis tracking), tiny sample.

Market-neutral SOL funding-carry: SHORT SOLUSDT.BINANCE USD-M perp (primary, receives 8h funding) + LONG equal-notional SOLUSDT.BINANCE_SPOT hedge (delta-neutral). Enters when 8h perp funding is clearly positive (>= entry_funding, +1bp/8h) and now holds through funding zero-crossings via a DEEP hysteresis band, flattening BOTH legs only when funding clearly inverts (<= exit_funding, -1bp/8h), a delta-neutral net-PnL basis-divergence stop (basis_stop_pct) trips, or a structural time cap is reached; a 3-bar minimum hold blocks sub-day round-trips. This makes one position ride an entire positive-funding regime, accumulating carry across many 8h periods so it amortizes the ~30bp four-fill round-trip. Uses only funding_rates supplementary data plus 8H OHLCV; goes flat (never a price-only fallback) when funding is unavailable within a 9h tolerance. Three core parameters (entry_funding, exit_funding, basis_stop_pct).

CLEAN RESTART 2026-09-04 — this run's verdict history and learning records were removed and it was restarted from verification. Its previous abandonment came from the pipeline, not from the market: the Layer-2 harness mis-bound @staticmethod helpers (fixed), QA issued terminal performance verdicts on an unoptimized smoke test (removed — QA now judges correctness only), and sandbox timeouts came from backtest-slot starvation (fixed). The hypothesis and the strategy code are unchanged. Verify the code on its merits; performance is decided later by the full backtest and the optimizer.

~111 traded days / 23 trades is a short, possibly SOL-spot-history-limited window for a carry with holds up to 90 days; per-trade variance (~±9% equity, skew -3.5, kurtosis 16.5) is large vs the tiny per-8h funding, so the smoke result can't distinguish 'no edge' from 'wrong window/untuned band'. Run full history, verify funding is credited, tune the funding band + basis_stop, require avg_trade_return_pct above the low-fee cross-venue floor before promoting.

Correct delta-neutral two-leg construction (spot BUY hedge submitted before perp SELL), funding-gated entry, deep hysteresis to reduce churn; low drawdown (3%) as expected for a hedged book

Funding is credited by the engine, so the backtest genuinely reflects the carry cash flow

Net loser after crediting funding: avg_trade_return_pct -0.211% (negative), profit_factor 0.95, expectancy -$220/trade, Sharpe -5.19 (CI [-7.13, -3.01]), PSR 0.0 — carry does not cover the 4-fill round-trip + basis tracking error

Tiny unmeasurable sample: 23 trades / 12 entries confined to 2020-09→2021-01 (data_days 111, only 332 bars jointly processed), cannot support walk-forward/holdout

Loses in all three vol-regime buckets; delta-neutral so benchmark_meaningful false, and absolute risk-adjusted metrics are deeply negative

Same negative-per-trade-carry outcome as the BTC spot-perp funding-carry sibling abandoned earlier today

~4mo (2020-2021)

multi-year

Verification failed (Layer 3 — sandbox backtest): No trades produced Bar type used: SOLUSDT.BINANCE-8-HOUR-LAST-EXTERNAL, Bars processed: 1000 Diagnostics: should_enter() returned a side 0 times over 885 evaluated bars -> your ENTRY CONDITION never triggered. Loosen the entry logic / thresholds. Ensure your strategy produces trades with the given data and parameters.

Verification failed (Layer 4 — QA review): - Fee-viability / carry-accumulation defect. The hypothesis's edge requires a MULTI-DAY hold accumulating many 8h funding periods so accrued carry clears the ~30bp round-trip cost (4 fills: perp 0.05%x2 + spot 0.10%x2, which the hypothesis itself acknowledges). But should_exit() flattens BOTH legs the moment funding <= exit_funding (0.0). On SOL, 8h funding dips non-positive frequently, so the book churns out after only ~4 funding periods (sandbox avg hold = 1d 6h), collecting a few bp of carry against a 30bp round-trip. The exit band (enter >=0.5bp, exit <=0) structurally prevents the accumulation the mechanism depends on. Confirmed by sandbox over 220 well-sampled trades: avg_trade_return_pct = 0.0002% (~0.2bp vs the 0.15% floor), total_return -9.71%, PF 0.876, Sharpe -0.35. - High turnover (77.4) and total_impact_usd $4,099 for a strategy with no net edge, driven by re-entry/exit churn around the funding zero-crossing.

Verification failed (Layer 4 — QA review): - Fee viability — now MEASURABLE and CONFIRMED negative. The iter-3 fix mechanically worked (holds stretched to 5d 5h, turnover 77→25.85, funding applied: 878 events), yet the strategy is still a net loser: avg_trade_return_pct = -0.0021% (negative per-trade expectancy), total_return -3.27%, PF 0.942, win_rate 0.50, over 72 reliable trades. Structural cause: the 4-fill round-trip is ~30bp (perp 0.05%×2 + SPOT 0.10%×2), but SOL's median 8h funding is ~1bp. Even the 5.2-day hold spans ~15.7 periods ≈ 15.7bp carry, ~14bp short of the 30bp round-trip; breakeven needs ~10 days of sustained positive funding but the band exits when funding inverts first.